Artificial Neural Network for Predicting Cardiovascular Autonomic Reflex Tests from Inflammatory Markers

Cardiac Autonomic Neuropathy (CAN) is a serious complication of diabetes that is associated with multi-organ complications, including cardiovascular, renal, and neurological complications. Cardiovascular Autonomic Reflex Tests (CARTs) are widely accepted as a gold standard measure of autonomic function to diagnose CAN. The aim of this paper is to predict the results of CARTs based on inflammatory biomarkers using a comprehensive dataset collected from a rural diabetes screening clinic at Charles Sturt University (CSU) (DiabHealth) with 2621 patient entries. An Artificial Neural Network (ANN) model optimized by the Sparse Categorical Cross Entropy Loss function is proposed to predict the CART results as normal, borderline, or abnormal. The ANN was compared with various baseline models, where it outperformed all with F1-values of 0.968, 0.904, 949, 0.949, and 0.926 for five autonomic function tests, being LS-HR, DB-HR, VA-HR, LS-BP, and HG-BP respectively. MCP-1, IGF-1, and IL-1Beta were found to be the most significant inflammatory markers for predicting CART results. Utilizing inflammatory markers from urine samples provides an accurate alternative opportunity for the identification of CAN and its progression, in addition to identifying possible treatment pathways based on inflammatory markers.